cuda-proj-tmpl

A minimal cmake based project skeleton for developping a CUDA application

Download this skeleton

git clone --recursive [email protected]:pkestene/cuda-proj-tmpl.git

Modern CMake and CUDA

See https://github.com/CLIUtils/modern_cmake

Requirements

• cmake version >= 3.18 (when using nvcc)
• cmake version >= 3.22 (when using nvc++/nvhpc compiler)
• cuda toolkit

How to build with nvcc ?

# set default CUDA flags passed to nvcc (Nvidia compiler wrapper)
# at least one hardware architecture flags
export CUDAFLAGS="-arch=sm_75 --expt-extended-lambda"
mkdir build
cd build
cmake ..
make
# then you can run the application
./src/saxpy_cuda

If you use cmake version >= 3.18, here is a slightly updated version using CMAKE_CUDA_ARCHITECTURE to pass information about actual GPU hardware target architecture:

# set default CUDA flags passed to nvcc (Nvidia compiler wrapper)
# at least one hardware architecture flags
export CUDAFLAGS="--expt-extended-lambda"
mkdir build
cd build
cmake -DCMAKE_CUDA_ARCHITECTURES="75" ..
make
# then you can run the application
./src/saxpy_cuda

How to build with nvc++ (from nvhpc) ?

# Warning, cmake 3.22.0 required
export CUDAFLAGS="--expt-extended-lambda"
export CXX=nvc++
mkdir build
cd build
cmake -DCMAKE_CUDA_HOST_COMPILER=nvc++ -DCMAKE_CUDA_ARCHITECTURES="75" ..
make
# then you can run the application
./src/saxpy_cuda